CLIMsystems Extreme Rainfall Climate Change Projection Solutions Glossary IDF – Intensity Duration Frequency DDF – Depth Duration Frequency GCM – Global Circulation Model RCM – Regional Climate Model RCP – Representative Concentration Pathway
Introduction Climate change changed precipitation characteristics in trend, variability and extremes, namely, annual total amount, seasonal patterns, and extremes. Changes in different aspects have specific impacts on natural and manmade systems. Table 1 presents a summary from annual to sub-hourly change in precipitation, potential implications, and CLIMsystems’ solutions to provide appropriate data for real applications. CLIMsystems’ solutions focus on real application and providing robust climate change data derived from our rich experience providing data and software service to different sectors over the last 15 years. Climate change impact/risk assessment needs highly specialized experience and knowledge of climate change science, and a thorough understanding of climate change data and its limitations and applicability to specific climate-related problems. Close and seamless cooperation of a team with experts and specialists in climate data and with sectoral experience can offer a range of integrated and useful information for different sectors. Common understanding of data applications, limitations and caveats, and transparent communication and interpretation of the data is very important in achieving credible outputs.
www.climsystems.com
1|Page
Figure 1: Climate data and analysis often needs to be processed to different levels of robustness depending on the nature of the project i.e. from screening to major infrastructure investment with either a long life cycle or critical lifelines function like a hospital or water/waste treatment facility. To assist our clients to better understand the implications of data and methods to extreme rainfall events analysis we have put together the following matrix and links to more detailed information. After digesting some or all of this information sometimes a quick call is helpful to further refine your specific methodological requirements. We are available for such consultations using a wide variety of technologies. Send us an email to arrange a call:
[email protected] or call +64 27 316 9777. Further information is available online from one of our Climate Scientists, Dr Chonghua Yin: https://www.linkedin.com/pulse/brief-introduction-how-apply-idf-information-chonghua-yin https://www.linkedin.com/pulse/small-note-updating-short-duration-idf-curves-under-climate-yin https://www.linkedin.com/pulse/do-get-confused-change-factor-method-bias-correction-when-yin https://www.linkedin.com/pulse/bias-correct-methods-used-statistical-adjustment-gcmrcmsdsm-yin https://www.linkedin.com/pulse/statistical-bias-correction-subdaily-daily-chonghua-yin https://www.linkedin.com/pulse/what-governs-decision-making-process-when-moving-from-chonghua-yin https://www.linkedin.com/pulse/analysis-extreme-precipitation-changing-climate-chonghua-yin https://www.linkedin.com/pulse/pathways-regionalized-climate-change-informaton-chonghua-yin
www.climsystems.com
2|Page
Table 1: Climate change related precipitation analysis and their application
Applications
Baseline and total seasonal and monthly changes in slow onset for water resources planning
Extreme changes, changes in return period, IDF, or DDF for water infrastructure design
Precipitation related climate change analysis type
Annual or monthly mean
Daily precipitation extremes
Recommended methodology
Change factor approach (Percentage change per degree, or percentage in different scenarios) Observation based Monthly historical data Multiple GCM and RCM monthly mean ensemble results
Generalized Extreme Value analysis, multiple distributions fitting testing.
CLIMsystems tool
SimCLIM monthly pattern scenario generator
SimCLIM GEV tool and inhouse tools
Potential linkage to other models
WEAP,DSSAT
Related infrastructure design
Historical data required GCM/RCM data required
www.climsystems.com
Daily observation time series Multiple GCM daily precipitation
Extreme changes in short duration (hourly), changes in return period, IDF, or DDF for urban water infrastructure design Sub-hourly to multiple day IDF or DDF for storm water system design
Water resources, flooding, inundation modelling
Extreme changes in short duration (hourly), changes in return period, IDF, or DDF for urban water infrastructure design
Hydrological model input data
Generalized Extreme Value analysis and IDF curve fitting, multiple distributions fitting testing. Subdaily observation data Multiple GCM/RCM 1 hourly or 3 hourly precipitation output Sub-hourly Subdaily extreme event analysis inhouse tool Related infrastructure design
Bias correction statistical downscaled with climate change projection
Very high resolution RCM precipitation changes for urban water systems 1-3 km resolution convection permitting RCM simulations
Subdaily or daily observations Multiple sources: subdaily or daily GCM or RCM
Sub-hourly precipitation observation RCM subhourly data
Multiple GCM daily/subdaily BCSD dataset SWAT, DHI, EWater, HECS,
WRF specific domain case by case
SWAT, DHI, EWater, HECS, SWMM, other 3|Page
Pros/Cons
Quick and easy to generate, but seasonal/monthly average change can’t reflect the extremes which is crucial for water resource management
models
models
Widely used for engineering design, applying daily GCM/RCM output may under estimate subdaily changes in extremes, which is important for flooding.
Widely used for engineering design. GCM/RCM sub-daily data reflects changes in higher temporal resolution. Not directly applicable for detailed flood modelling
SWMM, Flood Modeller Can be directly link to hydrological flood modelling and costeffectively for multiple scenario and multiple time slices.
Related infrastructure design models Can be directly link to hydrological flood modelling and may reflect some detailed change patterns. But it is costly and time consuming, therefor this method is not easy to apply to multiple scenario and time slices experiment
IDF/DDF Projection Methodology A warming climate might change the extreme precipitation quantiles represented by the DDF or IDF curves, emphasizing the need for updating the DDF or IDF curves used for the design of urban storm water management systems, including sewers, storm water management ponds, street curbs and gutters, catch basins, swales, among a significant variety of other types of infrastructure. Currently, DDF and IDF curves are usually developed using historical observed data with the assumption that the same underlying processes will govern future rainfall patterns and resulting DDF and IDF curves. This assumption is not valid under changing climatic conditions and therefore IDF curves that rely only on historical observations will misrepresent future conditions (Sugahara, et al. 2009; Milly et al. 2008). Global climate models (GCMs) provide understanding of climate change (i.e., non-stationary conditions) under different future emission scenarios, also known as representative concentration pathways (RCPs), and provide a way to update DDF and IDF curves under a changing climate. The updating procedure is illustrated in Figure 2 below.
www.climsystems.com
4|Page
Figure 2 Equidistance Quantile-Matching Method for generating future DDF and IDF curves under Climate Change
Historical DDF and IDF Curve (or table) Development A fundamental issue in the estimation of quantiles is the need to extrapolate to recurrence intervals significantly larger than the available records. This can be solved using regionalization, a standard practice for improving the estimation of event quantiles at sites with comparatively short records. In this report, the index regional flood frequency analysis method based on L-moments proposed by Hosking (Hosking and Wallis, 2005) is applied to estimate the regional rainfall quantiles at the site. Step 1: Derive the rainfall intensity time series of the different durations from historical hourly data time series. The durations include: 3, 6, 12, 24, 48, 72, 120, 144 and 168 hours, then select annual maxima series from the rainfall intensity series. Fit the annual maxima series for each duration to a group of probability distribution functions. Three distribution functions were tested, including: Generalized Logistic (GLO), GEV, and Gumbel distribution. L-moments method was deployed for distribution parameter estimation. Step 2: Assess their goodness of fit with the Anderson-Darling test to follow the method used (Vigione, 2008). The Anderson-Darling test measures the extent of the departure, in terms of probabilities, between a simulated hypothetical distribution and the frequency distribution for consideration. If the estimated probability is greater than some defined significance level, the test fails. In this case of the
www.climsystems.com
5|Page
three distributions, the GEV provided the best fit. Therefore the GEV parameters were used for further analysis. Step 3: Calculate rainfall depths for the range of return periods (including 2, 3, 5, 10, 15, 25, 50, 100, 200 and 300 year) for each storm duration using GEV distribution parameters obtained from Step 1. The values consist of the table of depth-duration-frequency (DDF). The intensity-frequency-duration (IDF) table can be computed directly from the DDF table by simply dividing the rainfall depths by duration in hours. Step 4: Generate the DDF and IDF curves based on the tables of DDF and IDF. A shape-preserving piecewise cubic interpolation is used to produce smooth DDF and IDF curves. Future DDF and IDF Curve (or table) Development The impacts of climate change on historical DDF and IDF are evaluated based on climate model data. In order to reduce uncertainty of climate change simulated by GCMs, the outputs of as many as possible of these GCMs are used. According to the Fifth Assessment Report of IPCC (AR5), there are 42 GCM models developed by various research centres around the world. Currently, this analysis adopts only 22 GCMs out of the 42 GCMs because: i) Not all the GCMs generate the two selected RCPs for future climate scenarios (i.e., 4.5 and 8.5); and ii) there are some technical issues related to downloading (such as connection to remote servers or repositories) for some GCM models. The basic procedure is employing an equidistant quantile matching (EQM) method to update the DDF and IDF curves under changing climate conditions. Step 1: GCM 3 hourly output for each grid cell were analysed using extreme value analysis (EVA) to calculate extreme rainfall amounts for current climate (called the baseline, 1986-2005), and the future periods of interest. For GCM resolution, please refer to the GCM summary table. Step 2: DDF and IDF change factors for the range of durations (3, 6, 12, 24, 48, and 72, 120, 144 and 168 hours) and return years (2, 3, 5, 10, 15, 25, 50, 100, 200 and 300 year) for each GCM under RCP4.5 and RCP8.5 were calculated by applying a pattern scaling approach (Li and Ye, 2011). Step 3: Interpolated the global DDF and IDF changes into the same spatial resolutions (0.5o x 0.5o) to construct a global database. Furthermore, a super ensemble method was carried out to derive ensemble statistics at different percentiles (e.g., 25th and 75th percentile of the GCM ensemble), with both RCP4.5 and RCP8.5 change factors for all GCMs being applied equally without any weighting. Step 4: Perturbed the historical estimated precipitation depth/intensity values of each duration and return period using the global DDF and IDF changes surrounding the site of XXXX. The global DDF and IDF changes show high variability around the world. There are also considerable differences among GCM members. To further reduce the impact of natural variabilities, the change factors applied in this work are averaged over the whole study area.
www.climsystems.com
6|Page
Step 5: DDF and IDF curves for selected future time periods (2020-2030, 2040-2050, 2070-2080, 20902100) were calculated using the same smoothing method as the historical ones.
A New Approach for Sub-Hourly Precipitation Climate Perturbation for Robust Water System Design CLIMsystems has developed equally efficient and scientifically credible approaches for hourly, daily and multiple day extreme rainfall perturbation for application in a variety of water-related projects where future climate-related time series data is critical. Sub-hourly extreme rainfall events have changed in magnitude more dramatically than daily, even hourly events in our climate changed world. This has put additional pressure on water infrastructure exemplified by increasing incidents of urban water management problems (loss of life and property, sewer system overtopping, infrastructure damage, disruption to services). Urban designers and engineers are developing and applying better testing and engineering solutions, however, because of constraints in modelling and data storage capacity, General Circulation Models (GCM) and Regional Climate Models (RCM) have until now not been adequately applied to provide data for sub-hourly water system modelling and adaptation decision support. To meet the growing necessity for better and more robust data to support decision making the team at CLIMsystems developed an innovative hybrid approach to perturb 5 to 15 minute (and other temporal resolution) time series data. The approach better reflects the climate change signal projected in GCMs and RCMs, including extremes, variation and total amount of change in precipitation that can be applied either for either a locale or spatially. There are several key factors accommodated in the approach that are important: (1) the method must preserve the extreme precipitation change information in the climate model (perturbed) output, especially the Intensity Duration Frequency (IDF) change characteristics; (2) the changes in non-extreme precipitation Cumulative Distribution Function (CDF) distribution in all quantiles must be retained; and, (3) the changes in annual and monthly precipitation must be retained. The approach devised and applied by CLIMsystems meets these three criteria. To avoid the problems of single model uncertainty and bias, multiple model ensembles and probability approaches are recommend by the Intergovernmental Panel on Climate Change (IPCC). With the recent and dramatic increase in the availability of GCM and RCM data plus advances in methods of analysis, an ensemble approach has become much more feasible. CLIMsystems holds extensive repositories of CMIP5 GCM and RCM data for the globe and through careful storage and post-processing the time and hence cost of generating scientifically rigorous sub-daily and daily outputs for flood and engineering studies has been dramatically reduced. This type of modelling and the data and methods required have already been successfully applied in several locations in the USA and Asia. The perturbed data has been used to stress test sewer system performance under various climate change scenarios. Our experience allows us to quickly and efficiently
www.climsystems.com
7|Page
generate useful data for flood modellers, system stress testers and planners looking at reducing climate risk through the design of infrastructure that considers the impacts of intense short duration rainfall. Table 2: Example of available CORDEX RCM data for North America* (CORDEX-NAM) with daily data
No.
GCM/RCM combination and resolution
No.
GCM/RCM combination and resolution
1 2
NAM-44-CanESM2-CRCM5 NAM-44-MPI-ESM-LR-CRCM5
9 10
NAM-44-HadGEM2-ES-RegCM4 NAM-44-MPI-ESM-LR-RegCM4
3 4
NAM-44-MPI-ESM-MR-CRCM5 NAM-44- CanESM2-CanRCM4
11 12
NAM-22-CanESM2-CanRCM4 NAM-22-GFDL-ESM2M-RegCM4
5 6
NAM-44-EC-EARTH-HIRAM5 NAM-44-CanESM2-RCA4
13 14
NAM-22-HadGEM2-ES-RegCM4 NAM-22-MPI-ESM-LR-RegCM4
7
NAM-44-EC-EARTH-RCA4
15
NAM-22-GFDL-ESM2M-WRF
8
NAM-44-GFDL-ESM2M-RegCM4
16
NAM-22-MPI-ESM-LR-WRF
Table 3: Available CORDEX-NAM hourly data No.
GCM-RCM combination
1
NAM-22-MPI-ESM-LR-RegCM4
2
NAM-44-CanESM2-CanRCM4
3
NAM-44-MPI-ESM-LR-RegCM4 Table 4: Available CMIP5 GCMs with 3 hourly precipitation data
1 2 3 4 5 6 7 8 9 10 11
GCM ACCESS1-0 BCC-CSM1-1 BCC-CSM1-1-M CCSM4 CMCC-CM CNRM-CM5 EC-EARTH FGOALS-g2 GFDL-CM3 GFDL-ESM2G GFDL-ESM2M
nlat nlon 145 192 64 128 160 320 192 288 240 480 128 256 160 320 60 128 90 144 90 144 90 144
12 13 14 15 16 17 18 19 20 21 22
GCM GISS-E2-H GISS-E2-R HadGEM2-ES INMCM4 IPSL-CM5A-LR IPSL-CM5A-MR MIROC5 MIROC-ESM MIROC-ESM-CHEM MRI-CGCM3 NorESM1-M
nlat nlon 90 144 90 144 145 192 120 180 96 96 143 144 128 256 64 128 64 128 160 320 96 144
*Other RCM data is available around the world. A complete list of RCM data can be found at: http://documents.climsystems.com/SimCLIM%204.0/CORDEX%20data%20availability%20in%20SimCLI M.pdf www.climsystems.com
8|Page
Sample Site 1: Washington DC precipitation IDF changes under RCP 8.5 Discussion points: For 3 hourly precipitation data only selected RCPs (RCP45 and RCP85) and time slices (20262045, 2081-2100), therefore only RCP85, 2090(2981-2100 data) were visualized This analysis only for demonstration or testing purpose, there was no pattern scaling or weighting method applied. The existing SimCLIM daily extreme precipitation patterns are obtained from pattern scaling method. Visualization and explanation need to be discussed. Characteristic of extreme precipitation change: (1) Short duration precipitation depth changes larger than longer duration. For example, 2 year return 3hr precipitation increase 25% in 2090, RCP85, while 168hr precipitation changes about 16%. (2) The longer return period event changes larger than shorter return period event. For example, 100 year 3 hr precipitation changes about 32%, while 3hr 2 year return event changes 25%. (3) Regional average shows a clear signal of above characteristics, but for individual grid cell, this may not clear hence the requirement to include a regional analysis.
www.climsystems.com
9|Page
Figure 3: Extreme precipitation changes in 2090 under RCP85, 22 GCM ensemble median, for study area (100W-55W, 36N-68N)
GCM data 22 GCM 3 hourly precipitation output for 1986-2005 (baseline), 2081-2100 period, RCP45 and RCP85.
Figure 4: 3 hour extreme precipitation analysis. 100 year return 3 hour precipitation is 2.6 inch during historical period. However in RCP8.5 scenario 2090, it will become 3.9 inch. Or this change can be expressed as: 2.7 inch becomes the 30 year return event.
www.climsystems.com
10 | P a g e
Sample site 2: 15 minutes precipitation time series data perturbation and depth-duration scaling
Figure 5: GCM 3 hourly data change factor ensemble result (8 GCMs).
Figure 6: 22 GCM ensemble extreme precipitation change factor scaling example results.
www.climsystems.com
11 | P a g e
Guidelines for Return Periods Ideally, the choice of a design return period should be based on an economic evaluation in which the costs of providing the drainage works are compared with the benefits derived. However, comprehensive local flood damage data are normally not available to the degree of precision required for cost-benefit analysis. For this reason, a general policy decision based on such considerations as land use, hazard to public safety and community expectations is more appropriate. Admittedly, for new drainage systems or drainage upgrading in some existing areas, particularly low lying ones or those in congested urban locations, the recommended standards may not be suitable or achievable. A pragmatic approach should be considered. In a case in Hong Kong the return periods recommended in urban drainage situations ranged from 50 to 200 years (Drainage Services Division, 2014). For the City of Dublin, Ireland a 100 year return period is applied for protection of flooding within properties (Greater Dublin Strategic Drainage Study, 2005). Guidelines for the selection of return period No. Type of project or feature
Return period (yr)
1 2 3 4 5 6 7 8 9 10 11 12
5 to 10 25 to 50 25 to 50 25 to 100 50 to 100 50 to 100 50 to 100 100 100 to 500 200 to 1000 100 to 10,000 (PMP) 10,000 (PMP)
Urban drainage [low risk] (up to 100 ha) Urban drainage [medium risk] (more than 100 ha) Road drainage Principal spillways (dams) Highway drainage Levees [medium risk] Urban drainage [high risk] (more than 1,000 ha) Flood plain development Bridge design (piers) Levees [high risk] Emergency spillways (dams) Freeboard hydrograph [for a class (c) dam]
Source: Ponce, V.M. Q & A on the return period to be used for design. Sourced 20 May 2016. http://returnperiod.sdsu.edu/
www.climsystems.com
12 | P a g e
References Anandhi, A., Frei, A., Pierson, D., Schneiderman, E., Zion, M., Lounsbury, D., Matonse, A. (2011). Examination of change factor methodologies for climate change impact assessment. Water Resources Research. 47 (3):1-10, W03501, doi:10.1029/2010WR009104. Dagnet, Y., Waskow, D., Elliot, C., Northrop, E., Thwaites, J., Modelgaard, K., Krnjaic, M., Levin, K., McGray, H. (2016). Staying on Track from Paris: Advancing the Key Elements of the Paris Agreement. Working Paper. Washington, DC: World Resources Institute. Available online at http//www.wri.org/ontrackfromparis. Dubai Meteorological Office (2016). Climate (Average Temperatures: 1984-2009; Precipitation: 19672009). Accessed 12 May 2016. http://web.archive.org/web/20131004223556/https://services.dubaiairports.ae/dubaimet/MET /Climate.aspx Dwyer, I.J., Reed, D.W., (1995). Allowance for discretization in hydrological and environmental risk estimation, Institute of Hydrology, Wallingford, UK. Eyre, J., Bhalchandra, P. (2014). Permeable Concrete Block Paving Applications in the United Arab Emirates. International Journal of Engineering Trends and Technology (IJETT) Vol 16 No 5: 227236. Francis, T. (2011).Extreme Precipitation Analysis at Sizewell: Final Report. Met Office. Hosking, J.R.M., Wallis, J. R., (2005). Regional Frequency Analysis: An approach Based on L-moments. Cambridge University Press, New York. Knutti et al. (2010). Good Practice Guidance Paper on Assessing and Combining Multi Model Climate Projections. IPCC Expert Meeting on Assessing and Combining Multi Model Climate Projections. National Center for Atmospheric Research Boulder, Colorado, USA25-27 January 2010. Liew, S.C., Raghavan, S.V., Liong, S. (2014). Development of Intensity-Duration-Frequency curves at ungauged sites: risk management under changing climate. Geoscience Letters, 1:8. Milly, P.C.D., Betancourt, J., Falkenmark, M., Hirsch, R.M., Zbigniew, W. Kundzewicz, D., Lettenmaier, P., Stouffer. R. J. (2008). Stationarity Is Dead: Whither Water Management? Science, 319 (5863), 573-574. [DOI:10.1126/science.1151915]. Mahoney, K., Alexander, M., Scott, J. D., & Barsugli, J. (2013). High-Resolution Downscaled Simulations of Warm-Season Extreme Precipitation Events in the Colorado Front Range under Past and Future Climates. Journal of Climate, 26(21), 8671-8689. Mamoon, A.A., Joergensen, N.E., Rahman, A., Qasam, H. (2014). Derivation of new design rainfall in Qatar using L-moment based index frequency approach. Gulf Organisation for Research and Development. International Journal of Sustainable Built Environment. Vol. 3: 111-118.
www.climsystems.com
13 | P a g e
Peck , A., Prodanovic , P., Slobodan P., Simonovic, P. (2012). Rainfall Intensity Duration Frequency Curves Under Climate Change: City of London, Ontario, Canada. Canadian Water Resources Journal / Revue Canadienne des Ressources Hydriques. Vol.37, Iss.3. Rudd, A. C., & Kay, A. L. (2015). Use of very high resolution climate model data for hydrological modelling: estimation of potential evaporation. Hydrology Research, nh2015028. Srivastav, R.K., Schardong, A. and Slobodan, S.P. (2014). Computerized Tool for the Development of Intensity-Duration-Frequency Curves under a Changing Climate: Technical Manual v.1 Water Resources Research Report no. 089, Facility for Intelligent Decision Support, Department of Civil and Environmental Engineering, London, Ontario, Canada, 52 pages. ISBN: (print) 978-0-77143087-9; (online) 978-0-7714-3088-6. Sherif, M., Chowdhury, R. (2014a). Analysis of rainfall, PMP and drought in the United Arab Emirates. International Journal of Climatology. March 2014. DOI: 10.1002/joc.3768. Sherif, M., Chowdhury, R. (2014b).Rainfall and Intensity-Duration-Frequency (IDF) Curves in the United Arab Emirates. May 2014. DOI: 10.1061/9780784413548.231. Sugahara, S.,I, Rocha R.P., Silveira R (2009). Non-stationary frequency analysis of extreme daily rainfall in Sao Paulo , Brazil, 29, 1339–1349. doi:10.1002/joc. Viglione, A, (2008). Contributed R-Package: nsRFA (Non-supervised Regional Frequency Analysis). URL: http://www.r-project.org. Wang, D., Hagen, S. C., & Alizad, K. (2013). Climate change impact and uncertainty analysis of extreme rainfall events in the Apalachicola River basin, Florida. Journal of Hydrology, 480, 125-135. WMO. (2009). Guide to Hydrological Practices, Volume II: Management of Water Resources and Application of Hydrological Practices, WMO-No. 168, 6th Edition, World Meteorological Organization. Yu, M., & Liu, Y. (2015). The possible impact of urbanization on a heavy rainfall event in Beijing. Journal of Geophysical Research: Atmospheres, 120 (16), 8132-8143.
www.climsystems.com
14 | P a g e
